feat: add solana-tx-landing skill

[abhinavgautam01]

Summary

Adds solana-tx-landing, a full novel skill submission for the Solana AI Kit skill bounty.

This is a production-oriented Claude Code / Codex skill for diagnosing and hardening Solana transaction landing flows. It helps agents investigate failed, expired, slow, dropped, or inconsistently confirmed transactions by looking across the full transaction lifecycle: blockhash fetch, transaction construction, simulation, wallet signing latency, send/rebroadcast, confirmation, priority fees, compute budget, RPC pool health, versioned transactions, address lookup tables, and Jito routing.

What is included

This PR adds a self-contained submission folder:

solana-tx-landing/
├── README.md
├── DEMO.md
├── LICENSE
├── install.sh
├── skill/
│   ├── SKILL.md
│   ├── agents/openai.yaml
│   └── references/
├── scripts/
├── tests/fixtures/
├── agents/
├── commands/
└── rules/

Key pieces:

• Progressive SKILL.md router with focused references for:
  • blockhash and confirmation
  • priority fees and compute budget
  • RPC health, retries, failover, minContextSlot, and maxRetries
  • simulation logs and program errors
  • wallet signing, versioned transactions, and address lookup tables
  • Jito bundles
  • mainnet readiness reporting
• Deterministic offline scanners:
  • scan_ts_transactions.py for TypeScript/JavaScript transaction send paths
  • parse_simulation_logs.py for simulation log classification
  • scan_anchor_compute.py for Rust/Anchor compute-risk patterns
  • diagnose_signature.py for saved getTransaction JSON and optional live read-only RPC diagnosis
  • tx_landing_report.py for combined transaction landing readiness reports
• MCP support:
  • mcp_server.py exposes the offline scanners to MCP-compatible agents
• Agent/operator surfaces:
  • command docs for diagnosis, hardening, and report workflows
  • focused agent instructions
  • safety rules
• Evidence-driven demo:
  • actual command outputs
  • before/after transaction sender story
  • conservative patch preview
  • temp-copy --fix verification
  • CI-gate example
  • MCP tool listing

Why this matters

Transaction landing is one of the most common production pain points for Solana builders. The same symptoms can come from very different causes: stale blockhashes, commitment mismatch, missing lastValidBlockHeight, skipped preflight, weak rebroadcast logic, public RPC endpoints, compute exhaustion, poor priority fee strategy, wallet signing latency, or real program/runtime errors.

This skill gives coding agents a deterministic way to separate those causes and propose concrete, minimal fixes without requiring private keys, signing, live transaction sending, or network access.

Safety model

• Offline by default
• No private-key handling
• No signing
• No live transaction sending
• Optional live signature diagnosis is read-only and requires an explicit RPC URL
• Static scanners are conservative and instruct agents to confirm findings against source code

Validation

Ran from solana-tx-landing/ in this PR branch:

PYTHONDONTWRITEBYTECODE=1 bash scripts/validate.sh

Output:

[OK] Skill frontmatter is valid
[OK] Reference routing is complete
[OK] Python scripts compile without bytecode artifacts
[OK] Scanner fixtures produce expected findings
[OK] Shell helper syntax and project-local install work
[OK] No generated Python artifacts found
[OK] Repository validation complete

Also validated the skill entry point:

PYTHONDONTWRITEBYTECODE=1 python3 /Users/gautam/.codex/skills/.system/skill-creator/scripts/quick_validate.py skill

Output:

Skill is valid!

Links

Standalone submission repo:

https://github.com/abhinavgautam01/solana-tx-landing-skill

Detailed demo:

https://github.com/abhinavgautam01/solana-tx-landing-skill/blob/main/DEMO.md

Related marketplace PR:

solanabr/skills#2

[abhinavgautam01]

ping @kauenet